Enterprise AI Adoption: Why 29% of Workers Sabotage Rollouts

A survey that reframes the AI rollout problem

Most debate about enterprise AI has focused on whether the technology works. New survey data suggests that question may be less important than whether organizations can absorb it. Writer's 2026 Enterprise AI Adoption Survey, run with Workplace Intelligence, polled 2,400 knowledge workers, split evenly between C-suite executives and employees 2. Its most striking finding is that 29% of employees admit to sabotaging their company's AI strategy, and that share rises to 44% among Gen Z workers 1.

The sabotage described is mostly passive or opportunistic rather than dramatic. It includes entering proprietary data into public AI tools, using unapproved software, deliberately producing low-quality outputs, refusing AI training, and manipulating performance metrics 1. Some of these behaviors look less like resistance to AI and more like resistance to how it is being imposed. Using a non-sanctioned tool, for instance, implies someone wants AI, just not the version IT selected.

Personal wins, organizational losses

Writer frames its findings as an "invisible crisis": AI is working for individuals but breaking organizations 2. The vendor reports that 97% benefit from AI personally, while only 23% of companies see significant return on investment 2. The fallout is cultural as well as financial. According to the survey, 56% of executives report power struggles, 54% say AI is fracturing their organization, and 60% plan to lay off employees who won't adopt it 2.

The figures shift slightly depending on who is reporting them. An analysis of the same survey from Scaling the Enterprise attributes the 97% benefit figure to executives rather than employees. It also splits ROI by technology: 29% of companies report significant returns from generative AI and 23% from AI agents 1. That analysis adds that 59% of companies spend more than $1 million a year on AI 1. These differences change the emphasis, not the conclusion. On either reading, the gap between enthusiasm and measurable organizational payoff is wide.

The layoff statistic deserves attention alongside the sabotage figure. When six in ten executives say they plan to fire non-adopters, covert pushback from employees looks less surprising. Threats of that kind tend to drive resistance underground rather than eliminate it.

Broken processes, not just reluctant people

A separate study points to a structural cause that the sabotage headline can hide. A Camunda-commissioned survey of 1,000 process leaders and 5,000 employees found that 72% of organizations say process-related problems have already caused AI initiatives to fail 3. Those failures cost an average of $1.55 million per business 3. Eighty-two percent of respondents believe AI investments will be a bust unless processes are re-engineered first 3.

The Camunda data also shows the same leadership-versus-staff split that Writer describes. Eighty-nine percent of organizations say AI is boosting team productivity, but only 64% of employees agree 3. Forty-four percent of employees have had to manually override AI outputs, and 46% say they would consider quitting over poorly implemented AI 3.

Read together, the two surveys suggest a feedback loop. AI gets layered onto workflows that were never redesigned. Employees spend time correcting its errors. Leadership reports productivity gains that staff don't experience. Frustration then shows up as workarounds, shadow tools, or deliberate underperformance.

Not a story of failure, exactly

All this sits awkwardly beside the popular claim, traced to an MIT study, that 95% of enterprise AI pilots fail. Scaling the Enterprise argues that enterprises are actually better at adopting AI than that figure implies. It points to a growing "AI elite" of companies pulling ahead as one of the most underreported stories in the space 1. Writer likewise markets its findings around what separates the 23% seeing returns from the 77% falling behind, highlighting governance and alignment between strategy, IT, and staff 2.

There is a caveat. Writer sells enterprise AI tools, and Camunda sells process orchestration software. Each survey frames the problem in a way that fits its sponsor's offering, so the numbers should be read with that in mind.

The takeaway

Even allowing for vendor framing, the surveys converge on one conclusion: AI's bottleneck in 2026 is organizational, not technical. The sabotage figure is better read as a symptom than a cause. It reflects rollouts that skip workflow redesign, overstate benefits from the top, and use job threats as an adoption strategy.

Companies seeing real returns appear to be the ones treating AI as a change-management and process problem rather than a software purchase. For everyone else, the 29% of employees quietly working against the plan may be the most honest signal of how it is actually going.